tracepipe 0.3.0__py3-none-any.whl → 0.3.2__py3-none-any.whl

tracepipe/__init__.py

@@ -81,7 +81,7 @@ from .core import TracePipeConfig, TracePipeMode
 from .snapshot import DiffResult, Snapshot, diff, snapshot
 
 # === VERSION ===
-__version__ = "0.3.0"
+__version__ = "0.3.2"
 
 # === MINIMAL __all__ ===
 __all__ = [

tracepipe/api.py

@@ -75,12 +75,18 @@ def enable(
         # Custom configuration
         tp.enable(mode="ci", merge_provenance=True)
     """
+    ctx = get_context()
+
+    # If already enabled, reset accumulated state to prevent duplicate warnings/stats
+    # This handles the common case of re-running scripts in notebooks/IDEs
+    if ctx.enabled:
+        _reset_accumulated_state(ctx)
+
     # Get or create config
     # If config is provided explicitly, use it
     # Otherwise, start with existing context config (if any) or create new default
     if config is None:
-        existing_ctx = get_context()
-        config = existing_ctx.config  # Use existing config as base
+        config = ctx.config  # Use existing config as base
 
     # Handle mode
     if mode is not None:
@@ -115,14 +121,14 @@ def enable(
         ctx = TracePipeContext(config=config, backend=backend, identity=identity)
         set_context(ctx)
     else:
-        ctx = get_context()
         ctx.config = config
         # Also update config in row_manager and store (they may have their own references)
         ctx.row_manager.config = config
         ctx.store.config = config
 
-    # Add watched columns
+    # Add watched columns (reset first if re-enabling to avoid stale watches)
     if watch:
+        ctx.watched_columns.clear()
         ctx.watched_columns.update(watch)
 
     if not ctx.enabled:
@@ -132,6 +138,50 @@ def enable(
     return _get_module()
 
 
+def _reset_accumulated_state(ctx: TracePipeContext) -> None:
+    """
+    Reset accumulated lineage state without disabling instrumentation.
+
+    Called when enable() is invoked on an already-enabled context to prevent
+    state accumulation across multiple script runs in the same Python process.
+    """
+    store = ctx.store
+
+    # Clear merge stats (prevents duplicate warnings)
+    if hasattr(store, "merge_stats"):
+        store.merge_stats.clear()
+
+    # Clear bulk drops
+    if hasattr(store, "bulk_drops"):
+        store.bulk_drops.clear()
+
+    # Clear steps
+    if hasattr(store, "_steps"):
+        store._steps.clear()
+
+    # Clear in-memory diffs
+    if hasattr(store, "_clear_in_memory"):
+        store._clear_in_memory()
+
+    # Reset step counter
+    if hasattr(store, "_step_counter"):
+        store._step_counter = 0
+
+    # Clear merge mappings
+    if hasattr(store, "merge_mappings"):
+        store.merge_mappings.clear()
+
+    # Clear aggregation mappings
+    if hasattr(store, "aggregation_mappings"):
+        store.aggregation_mappings.clear()
+
+    # Reset row identity manager
+    ctx.row_manager.clear()
+
+    # Clear watched columns (will be re-added if watch param provided)
+    ctx.watched_columns.clear()
+
+
 def disable() -> types.ModuleType:
     """
     Disable TracePipe and restore original pandas methods.

tracepipe/convenience.py

@@ -385,22 +385,27 @@ def check(
                 )
             )
 
-        if stats.left_dup_rate > 0.01:
+        # Note on dup_rate semantics:
+        # - left_dup_rate = fraction of LEFT rows appearing >1 times in result
+        #   This happens when RIGHT table has duplicate join keys
+        # - right_dup_rate = fraction of RIGHT rows appearing >1 times in result
+        #   This happens when LEFT table has duplicate join keys
+        if stats.right_dup_rate > 0.01:
             warnings_list.append(
                 CheckWarning(
                     category="duplicate_keys",
                     severity="fact",
-                    message=f"Left table has {stats.left_dup_rate:.1%} duplicate join keys",
-                    details={"step_id": step_id, "dup_rate": stats.left_dup_rate},
+                    message=f"Left table has {stats.right_dup_rate:.1%} duplicate join keys",
+                    details={"step_id": step_id, "dup_rate": stats.right_dup_rate},
                 )
             )
-        if stats.right_dup_rate > 0.01:
+        if stats.left_dup_rate > 0.01:
             warnings_list.append(
                 CheckWarning(
                     category="duplicate_keys",
                     severity="fact",
-                    message=f"Right table has {stats.right_dup_rate:.1%} duplicate join keys",
-                    details={"step_id": step_id, "dup_rate": stats.right_dup_rate},
+                    message=f"Right table has {stats.left_dup_rate:.1%} duplicate join keys",
+                    details={"step_id": step_id, "dup_rate": stats.left_dup_rate},
                 )
             )
 
@@ -733,9 +738,14 @@ def _build_trace_result(row_id: int, ctx, include_ghost: bool) -> TraceResult:
     store = ctx.store
 
     drop_event = store.get_drop_event(row_id)
-    history = store.get_row_history(row_id)
     merge_origin = store.get_merge_origin(row_id)
 
+    # Use lineage-aware history to include pre-merge parent events
+    if hasattr(store, "get_row_history_with_lineage"):
+        history = store.get_row_history_with_lineage(row_id)
+    else:
+        history = store.get_row_history(row_id)
+
     dropped_at = None
     if drop_event:
         dropped_at = {

tracepipe/storage/lineage_store.py

@@ -485,6 +485,9 @@ class InMemoryLineageStore:
 
         CONTRACT: Returned list has monotonically increasing step_id.
         Convenience layer may reverse for display.
+
+        Note: This returns only direct events for this row_id.
+        Use get_row_history_with_lineage() to include pre-merge parent history.
         """
         step_map = {s.step_id: s for s in self._steps}
         events = []
@@ -546,6 +549,65 @@ class InMemoryLineageStore:
 
         return events
 
+    def get_row_history_with_lineage(self, row_id: int, max_depth: int = 10) -> list[dict]:
+        """
+        Get row history including pre-merge parent history.
+
+        Follows merge lineage recursively to build complete cell provenance.
+        This is essential for tracking changes that happened before merge operations.
+
+        Args:
+            row_id: Row ID to trace
+            max_depth: Maximum merge depth to follow (prevents infinite loops)
+
+        Returns:
+            List of events in chronological order, including parent row events.
+        """
+        visited: set[int] = set()
+
+        def _collect_history(rid: int, depth: int) -> list[dict]:
+            if depth > max_depth or rid in visited:
+                return []
+            visited.add(rid)
+
+            events = []
+
+            # Check if this row came from a merge
+            origin = self.get_merge_origin(rid)
+            if origin and origin["left_parent"] is not None:
+                # Recursively get parent's history first (chronological order)
+                parent_events = _collect_history(origin["left_parent"], depth + 1)
+                events.extend(parent_events)
+
+            # Add this row's direct events
+            events.extend(self.get_row_history(rid))
+
+            return events
+
+        all_events = _collect_history(row_id, 0)
+
+        # Sort by step_id to ensure chronological order across lineage
+        all_events.sort(key=lambda e: e["step_id"])
+
+        return all_events
+
+    def get_cell_history_with_lineage(
+        self, row_id: int, column: str, max_depth: int = 10
+    ) -> list[dict]:
+        """
+        Get cell history for a specific column, including pre-merge parent history.
+
+        Args:
+            row_id: Row ID to trace
+            column: Column name to filter events for
+            max_depth: Maximum merge depth to follow
+
+        Returns:
+            List of events for this column in chronological order.
+        """
+        all_events = self.get_row_history_with_lineage(row_id, max_depth)
+        return [e for e in all_events if e["col"] == column]
+
     def get_dropped_rows(self, step_id: Optional[int] = None) -> list[int]:
         """Get all dropped row IDs, optionally filtered by step."""
         if step_id is not None:
@@ -648,7 +710,7 @@ class InMemoryLineageStore:
         diffs = list(self._iter_all_diffs())
 
         data = {
-            "tracepipe_version": "0.3.0",
+            "tracepipe_version": "0.3.2",
             "export_timestamp": time.time(),
             "total_diffs": len(diffs),
             "total_steps": len(self._steps),

tracepipe/value_provenance.py

@@ -19,7 +19,6 @@ from typing import Any, Optional
 import pandas as pd
 
 from .context import get_context
-from .core import ChangeType
 
 
 @dataclass
@@ -96,7 +95,12 @@ class ValueHistory:
         }
 
 
-def explain_value(row_id: int, column: str, df: Optional[pd.DataFrame] = None) -> ValueHistory:
+def explain_value(
+    row_id: int,
+    column: str,
+    df: Optional[pd.DataFrame] = None,
+    follow_lineage: bool = True,
+) -> ValueHistory:
     """
     Get complete history of a specific cell's value.
 
@@ -104,6 +108,7 @@ def explain_value(row_id: int, column: str, df: Optional[pd.DataFrame] = None) -
         row_id: Row ID to trace
         column: Column name
         df: Optional DataFrame for current value lookup
+        follow_lineage: If True, include pre-merge parent history (default: True)
 
     Returns:
         ValueHistory with all changes to this cell
@@ -121,35 +126,38 @@ def explain_value(row_id: int, column: str, df: Optional[pd.DataFrame] = None) -
             if len(matches) > 0 and column in df.columns:
                 current_value = df.iloc[matches[0]][column]
 
-    # Collect all events for this cell
+    # Collect events - use lineage-aware method if requested
+    if follow_lineage and hasattr(store, "get_cell_history_with_lineage"):
+        # Get cell history including pre-merge parent history
+        raw_events = store.get_cell_history_with_lineage(row_id, column)
+    else:
+        # Fallback to direct row_id lookup only
+        raw_events = [e for e in store.get_row_history(row_id) if e["col"] == column]
+
+    # Convert to ValueEvent objects
     events = []
-    step_map = {s.step_id: s for s in store.steps}
     became_null_at = None
     became_null_by = None
 
-    for diff in store._iter_all_diffs():
-        if diff["row_id"] == row_id and diff["col"] == column:
-            step = step_map.get(diff["step_id"])
-
-            events.append(
-                ValueEvent(
-                    step_id=diff["step_id"],
-                    operation=step.operation if step else "unknown",
-                    old_value=diff["old_val"],
-                    new_value=diff["new_val"],
-                    change_type=ChangeType(diff["change_type"]).name,
-                    timestamp=step.timestamp if step else 0,
-                    code_location=(
-                        f"{step.code_file}:{step.code_line}" if step and step.code_file else None
-                    ),
-                )
+    for diff in raw_events:
+        events.append(
+            ValueEvent(
+                step_id=diff["step_id"],
+                operation=diff.get("operation", "unknown"),
+                old_value=diff["old_val"],
+                new_value=diff["new_val"],
+                change_type=diff.get("change_type", "UNKNOWN"),
+                timestamp=diff.get("timestamp", 0) or 0,
+                code_location=diff.get("code_location"),
             )
+        )
 
-            # Track when value became null
-            if became_null_at is None and pd.isna(diff["new_val"]) and not pd.isna(diff["old_val"]):
-                became_null_at = diff["step_id"]
-                became_null_by = step.operation if step else "unknown"
+        # Track when value became null
+        if became_null_at is None and pd.isna(diff["new_val"]) and not pd.isna(diff["old_val"]):
+            became_null_at = diff["step_id"]
+            became_null_by = diff.get("operation", "unknown")
 
+    # Events should already be sorted by step_id from lineage method
     events.sort(key=lambda e: e.step_id)
 
     return ValueHistory(

tracepipe-0.3.2.dist-info/METADATA

@@ -0,0 +1,308 @@
+Metadata-Version: 2.4
+Name: tracepipe
+Version: 0.3.2
+Summary: Row-level data lineage tracking for pandas pipelines
+Project-URL: Homepage, https://github.com/tracepipe/tracepipe
+Project-URL: Documentation, https://tracepipe.github.io/tracepipe/
+Project-URL: Repository, https://github.com/tracepipe/tracepipe.git
+Project-URL: Issues, https://github.com/tracepipe/tracepipe/issues
+Project-URL: Changelog, https://tracepipe.github.io/tracepipe/changelog/
+Author: Gauthier Piarrette
+License: MIT License
+        
+        Copyright (c) 2026 Gauthier Piarrette
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+License-File: LICENSE
+Keywords: data-engineering,data-lineage,data-quality,debugging,observability,pandas
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Developers
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Requires-Python: >=3.9
+Requires-Dist: numpy>=1.20.0
+Requires-Dist: pandas>=1.5.0
+Provides-Extra: all
+Requires-Dist: psutil>=5.9.0; extra == 'all'
+Requires-Dist: pyarrow>=10.0.0; extra == 'all'
+Provides-Extra: arrow
+Requires-Dist: pyarrow>=10.0.0; extra == 'arrow'
+Provides-Extra: dev
+Requires-Dist: black>=23.0.0; extra == 'dev'
+Requires-Dist: pre-commit>=3.5.0; extra == 'dev'
+Requires-Dist: pytest-cov>=4.0.0; extra == 'dev'
+Requires-Dist: pytest>=7.0.0; extra == 'dev'
+Requires-Dist: ruff>=0.1.0; extra == 'dev'
+Requires-Dist: taskipy>=1.12.0; extra == 'dev'
+Provides-Extra: docs
+Requires-Dist: mkdocs-material>=9.5.0; extra == 'docs'
+Requires-Dist: mkdocs>=1.5.0; extra == 'docs'
+Requires-Dist: mkdocstrings[python]>=0.24.0; extra == 'docs'
+Requires-Dist: pymdown-extensions>=10.0.0; extra == 'docs'
+Provides-Extra: memory
+Requires-Dist: psutil>=5.9.0; extra == 'memory'
+Description-Content-Type: text/markdown
+
+<div align="center">
+
+# TracePipe
+
+### Row-level data lineage for pandas pipelines
+
+**Know exactly where every row went, why values changed, and how your data transformed.**
+
+[![PyPI version](https://img.shields.io/pypi/v/tracepipe.svg)](https://pypi.org/project/tracepipe/)
+[![Python 3.9+](https://img.shields.io/pypi/pyversions/tracepipe.svg)](https://pypi.org/project/tracepipe/)
+[![CI](https://github.com/gauthierpiarrette/tracepipe/actions/workflows/ci.yml/badge.svg)](https://github.com/gauthierpiarrette/tracepipe/actions/workflows/ci.yml)
+[![codecov](https://codecov.io/gh/gauthierpiarrette/tracepipe/branch/main/graph/badge.svg)](https://codecov.io/gh/gauthierpiarrette/tracepipe)
+[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
+[![Docs](https://img.shields.io/badge/docs-mkdocs-blue.svg)](https://gauthierpiarrette.github.io/tracepipe/)
+
+[Getting Started](#getting-started) · [Documentation](https://gauthierpiarrette.github.io/tracepipe/) · [Examples](#real-world-example)
+
+</div>
+
+---
+
+## Why TracePipe?
+
+Data pipelines are black boxes. Rows vanish. Values change. You're left guessing.
+
+```python
+df = pd.read_csv("customers.csv")
+df = df.dropna()                      # Some rows disappear
+df = df.merge(regions, on="zip")      # New rows appear, some vanish
+df["income"] = df["income"].fillna(0) # Values change silently
+df = df[df["age"] >= 18]              # More rows gone
+# What happened to customer C-789? 🤷
+```
+
+**TracePipe gives you the complete audit trail — zero code changes required.**
+
+---
+
+## Getting Started
+
+```bash
+pip install tracepipe
+```
+
+```python
+import tracepipe as tp
+import pandas as pd
+
+tp.enable(mode="debug", watch=["income"])
+
+df = pd.read_csv("customers.csv")
+df = df.dropna()
+df["income"] = df["income"].fillna(0)
+df = df[df["age"] >= 18]
+
+tp.check(df)  # See what happened
+```
+
+```
+TracePipe Check: [OK] Pipeline healthy
+
+Retention: 847/1000 (84.7%)
+Dropped: 153 rows
+  • DataFrame.dropna: 42
+  • DataFrame.__getitem__[mask]: 111
+
+Value changes: 23 cells modified
+  • DataFrame.fillna: 23 (income)
+```
+
+That's it. **One import, full visibility.**
+
+---
+
+## Core API
+
+| Function | What it does |
+|----------|--------------|
+| `tp.enable()` | Start tracking |
+| `tp.check(df)` | Health check — retention, drops, changes |
+| `tp.trace(df, where={"id": "C-789"})` | Follow a row's complete journey |
+| `tp.why(df, col="income", row=5)` | Explain why a cell has its current value |
+| `tp.report(df, "audit.html")` | Export interactive HTML report |
+
+---
+
+## Key Features
+
+<table>
+<tr>
+<td width="50%">
+
+### 🔍 Zero-Code Instrumentation
+TracePipe patches pandas at runtime. Your existing code works unchanged.
+
+### 📊 Complete Provenance
+Track drops, transforms, merges, and cell-level changes with before/after values.
+
+</td>
+<td width="50%">
+
+### 🎯 Business-Key Lookups
+Find rows by their values: `tp.trace(df, where={"email": "alice@example.com"})`
+
+### ⚡ Production-Ready
+1.0-2.8x overhead (varies by operation). Tested on DataFrames up to 1M rows.
+
+</td>
+</tr>
+</table>
+
+---
+
+## Real-World Example
+
+```python
+import tracepipe as tp
+import pandas as pd
+
+tp.enable(mode="debug", watch=["age", "income", "label"])
+
+# Load and clean
+df = pd.read_csv("training_data.csv")
+df = df.dropna(subset=["label"])
+df["income"] = df["income"].fillna(df["income"].median())
+df = df[df["age"] >= 18]
+
+# Audit
+print(tp.check(df))
+```
+
+```
+Retention: 8234/10000 (82.3%)
+Dropped: 1766 rows
+  • DataFrame.dropna: 423
+  • DataFrame.__getitem__[mask]: 1343
+
+Value changes: 892 cells
+  • DataFrame.fillna: 892 (income)
+```
+
+```python
+# Why does this customer have a filled income?
+tp.why(df, col="income", where={"customer_id": "C-789"})
+```
+
+```
+Cell History: row 156, column 'income'
+  Current value: 45000.0
+  [i] Was null at step 1 (later recovered)
+
+  History (1 change):
+    None -> 45000.0
+      by: DataFrame.fillna
+```
+
+---
+
+## Two Modes
+
+| Mode | Use Case | What's Tracked |
+|------|----------|----------------|
+| **CI** (default) | Production pipelines | Step counts, retention rates, merge warnings |
+| **Debug** | Development | Full row history, cell diffs, merge parents, group membership |
+
+```python
+tp.enable(mode="ci")     # Lightweight
+tp.enable(mode="debug")  # Full lineage
+```
+
+---
+
+## What's Tracked
+
+| Operation | Coverage |
+|-----------|----------|
+| `dropna`, `drop_duplicates`, `query`, `df[mask]` | ✅ Full |
+| `fillna`, `replace`, `loc[]=`, `iloc[]=` | ✅ Full (cell diffs) |
+| `merge`, `join` | ✅ Full (parent tracking) |
+| `groupby().agg()` | ✅ Full (group membership) |
+| `sort_values`, `head`, `tail`, `sample` | ✅ Full |
+| `apply`, `pipe` | ⚠️ Partial |
+
+---
+
+## Data Quality Contracts
+
+```python
+(tp.contract()
+    .expect_unique("customer_id")
+    .expect_no_nulls("email")
+    .expect_retention(min_rate=0.9)
+    .check(df)
+    .raise_if_failed())
+```
+
+---
+
+## Documentation
+
+📚 **[Full Documentation](https://gauthierpiarrette.github.io/tracepipe/)**
+
+- [Quickstart](https://gauthierpiarrette.github.io/tracepipe/getting-started/quickstart/)
+- [User Guide](https://gauthierpiarrette.github.io/tracepipe/guide/concepts/)
+- [API Reference](https://gauthierpiarrette.github.io/tracepipe/api/)
+- [Examples](https://gauthierpiarrette.github.io/tracepipe/examples/ml-pipeline/)
+
+---
+
+## Contributing
+
+```bash
+git clone https://github.com/gauthierpiarrette/tracepipe.git
+cd tracepipe
+pip install -e ".[dev]"
+pytest tests/ -v
+```
+
+See [CONTRIBUTING](https://gauthierpiarrette.github.io/tracepipe/contributing/) for guidelines.
+
+---
+
+## License
+
+MIT License. See [LICENSE](LICENSE).
+
+---
+
+<div align="center">
+
+**Stop guessing where your rows went.**
+
+```bash
+pip install tracepipe
+```
+
+⭐ Star us on GitHub if TracePipe helps your data work!
+
+</div>

{tracepipe-0.3.0.dist-info → tracepipe-0.3.2.dist-info}/RECORD

@@ -1,13 +1,13 @@
-tracepipe/__init__.py,sha256=ZO6-yKMpguohwQLSRovuJoakb7kN1ZveSBwlGwhC-ho,3342
-tracepipe/api.py,sha256=KFO0NYRaGqRevbNyFSCFK4ryhFwdixFtUnTeNabwb6o,11862
+tracepipe/__init__.py,sha256=MuwxV2mU4XxHqab62vQxaDAlhMvRCgUCmr_YU9R16ss,3342
+tracepipe/api.py,sha256=WdcKvvzI3voDt6fxZWa8vjyZQU8lfRshx7T78oj7oFE,13351
 tracepipe/context.py,sha256=_povLpqa5wd_ESHt5hbSmWTSMTF3nUfeutEQo4RMK2E,3856
 tracepipe/contracts.py,sha256=m-rjPrgnCiAgKEkweOS7P95jrjDptt5UPdvUlqaV_rU,16226
-tracepipe/convenience.py,sha256=9F4rLx7AGWwNPKhuJMZD-6PG-QiZq0_mzfmnoU28x6U,26036
+tracepipe/convenience.py,sha256=SZGcSOKPjAeJ9udPP_Fa_zTZY5GeDX61W6uftMwafjc,26563
 tracepipe/core.py,sha256=kAXks694rR0Z4tD7Gyty0TyJGWx2whsSdteYYpHuazo,8010
 tracepipe/debug.py,sha256=6t2GKVZLwn7SJLhrStE9qsmTiVIHATTE3jJPQ2DYtnc,10140
 tracepipe/safety.py,sha256=jTBZv4QGDJfnZETsSZeMKbdOUtGXk-_XkmllhnGWM-M,5537
 tracepipe/snapshot.py,sha256=OLREzE1_LkWITluG_Bqeb7Y4pAKb8Lb3zJEF3cxnloU,13967
-tracepipe/value_provenance.py,sha256=cCNDvMduYiFkTzfam5EpBNZI54RL4OtMLP6xNaM00ec,9092
+tracepipe/value_provenance.py,sha256=ogky6aOaZ-6K2uNBQxlXpmCeuvK434Hisj30zesRTd8,9330
 tracepipe/instrumentation/__init__.py,sha256=pd0n6Z9m_V3gcBv097cXWFOZEzAP9sAq1jjQnNRrDZ8,222
 tracepipe/instrumentation/apply_capture.py,sha256=cMThWzNXqWQENuMrCGTne1hO6fqaQFV7zJYNpsPTW4w,14463
 tracepipe/instrumentation/filter_capture.py,sha256=onlYLU5bBZSM3WmxM2AFHfktnlx7ReG-brEn5eZ_N10,15830
@@ -17,13 +17,13 @@ tracepipe/instrumentation/pandas_inst.py,sha256=2YSoju9ml2PjLOYzsx8MHH1iqhjgnXHb
 tracepipe/instrumentation/series_capture.py,sha256=N1Cf-pQDh23qQLLd8DNsxbcaD-91sTJkRd5AnccKZGE,10649
 tracepipe/storage/__init__.py,sha256=pGFMfbIgIi2kofVPwYDqe2HTYMYJoabiGjTq77pYi-g,348
 tracepipe/storage/base.py,sha256=7DV_-rp37DjBMr9B1w85hLVYhC8OQShk2PcEhT-n4tE,4894
-tracepipe/storage/lineage_store.py,sha256=KPN-OZOgkZeiIptodQst-Obp9krcuE7Erpc9NX53jKw,25148
+tracepipe/storage/lineage_store.py,sha256=swMMf59isoCQZHaezCmquA-0R5iGNH3eGWjc9d9LGmo,27392
 tracepipe/storage/row_identity.py,sha256=HBU0gTTJlFtFTcAdUCKuX-c9cHa0lo3CDIodDPDgOzA,17161
 tracepipe/utils/__init__.py,sha256=CI_GXViCjdMbu1j6HuzZhoQZEW0sIB6WAve6j5pfOC0,182
 tracepipe/utils/value_capture.py,sha256=wGgegQmJnVHxHbwHSH9di7JAOBChzD3ERJrabZNiayk,4092
 tracepipe/visualization/__init__.py,sha256=M3s44ZTUNEToyghjhQW0FgbmWHKPr4Xc-7iNF6DpI_E,132
 tracepipe/visualization/html_export.py,sha256=G0hfZTJctUCfpun17zXX1NIXhvJZbca6hKmP3rcIjbg,42282
-tracepipe-0.3.0.dist-info/METADATA,sha256=oEiGG2V8ya2J3ZKYU_oAfLIqYrZdgwqBRaKup44U-Uw,15478
-tracepipe-0.3.0.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
-tracepipe-0.3.0.dist-info/licenses/LICENSE,sha256=HMOAFHBClL79POwWL-2_aDcx42DJAq7Ce-nwJPvMB9U,1075
-tracepipe-0.3.0.dist-info/RECORD,,
+tracepipe-0.3.2.dist-info/METADATA,sha256=ik5FLmADKLqj25TprTnJPi21SW4EJ88mBTG-aQ4p-gc,9152
+tracepipe-0.3.2.dist-info/WHEEL,sha256=WLgqFyCfm_KASv4WHyYy0P3pM_m7J5L9k2skdKLirC8,87
+tracepipe-0.3.2.dist-info/licenses/LICENSE,sha256=HMOAFHBClL79POwWL-2_aDcx42DJAq7Ce-nwJPvMB9U,1075
+tracepipe-0.3.2.dist-info/RECORD,,

tracepipe-0.3.0.dist-info/METADATA

@@ -1,575 +0,0 @@
-Metadata-Version: 2.4
-Name: tracepipe
-Version: 0.3.0
-Summary: Row-level data lineage tracking for pandas pipelines
-Project-URL: Homepage, https://github.com/tracepipe/tracepipe
-Project-URL: Documentation, https://tracepipe.github.io/tracepipe/
-Project-URL: Repository, https://github.com/tracepipe/tracepipe.git
-Project-URL: Issues, https://github.com/tracepipe/tracepipe/issues
-Project-URL: Changelog, https://tracepipe.github.io/tracepipe/changelog/
-Author: Gauthier Piarrette
-License: MIT License
-        
-        Copyright (c) 2026 Gauthier Piarrette
-        
-        Permission is hereby granted, free of charge, to any person obtaining a copy
-        of this software and associated documentation files (the "Software"), to deal
-        in the Software without restriction, including without limitation the rights
-        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
-        copies of the Software, and to permit persons to whom the Software is
-        furnished to do so, subject to the following conditions:
-        
-        The above copyright notice and this permission notice shall be included in all
-        copies or substantial portions of the Software.
-        
-        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
-        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
-        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
-        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
-        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
-        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
-        SOFTWARE.
-License-File: LICENSE
-Keywords: data-engineering,data-lineage,data-quality,debugging,observability,pandas
-Classifier: Development Status :: 4 - Beta
-Classifier: Intended Audience :: Developers
-Classifier: Intended Audience :: Science/Research
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.9
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Topic :: Scientific/Engineering
-Classifier: Topic :: Software Development :: Libraries :: Python Modules
-Requires-Python: >=3.9
-Requires-Dist: numpy>=1.20.0
-Requires-Dist: pandas>=1.5.0
-Provides-Extra: all
-Requires-Dist: psutil>=5.9.0; extra == 'all'
-Requires-Dist: pyarrow>=10.0.0; extra == 'all'
-Provides-Extra: arrow
-Requires-Dist: pyarrow>=10.0.0; extra == 'arrow'
-Provides-Extra: dev
-Requires-Dist: black>=23.0.0; extra == 'dev'
-Requires-Dist: pre-commit>=3.5.0; extra == 'dev'
-Requires-Dist: pytest-cov>=4.0.0; extra == 'dev'
-Requires-Dist: pytest>=7.0.0; extra == 'dev'
-Requires-Dist: ruff>=0.1.0; extra == 'dev'
-Requires-Dist: taskipy>=1.12.0; extra == 'dev'
-Provides-Extra: docs
-Requires-Dist: mkdocs-material>=9.5.0; extra == 'docs'
-Requires-Dist: mkdocs>=1.5.0; extra == 'docs'
-Requires-Dist: mkdocstrings[python]>=0.24.0; extra == 'docs'
-Requires-Dist: pymdown-extensions>=10.0.0; extra == 'docs'
-Provides-Extra: memory
-Requires-Dist: psutil>=5.9.0; extra == 'memory'
-Description-Content-Type: text/markdown
-
-# TracePipe
-
-**Row-level data lineage for pandas pipelines.**
-
-TracePipe automatically tracks what happens to every row and cell in your DataFrame — drops, transformations, merges, and value changes. Zero code changes required.
-
-[![PyPI version](https://img.shields.io/pypi/v/tracepipe.svg)](https://pypi.org/project/tracepipe/)
-[![Python 3.9+](https://img.shields.io/pypi/pyversions/tracepipe.svg)](https://pypi.org/project/tracepipe/)
-[![CI](https://github.com/gauthierpiarrette/tracepipe/actions/workflows/ci.yml/badge.svg)](https://github.com/gauthierpiarrette/tracepipe/actions/workflows/ci.yml)
-[![codecov](https://codecov.io/gh/gauthierpiarrette/tracepipe/branch/main/graph/badge.svg)](https://codecov.io/gh/gauthierpiarrette/tracepipe)
-[![License: MIT](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
-[![Documentation](https://img.shields.io/badge/docs-mkdocs-blue.svg)](https://gauthierpiarrette.github.io/tracepipe/)
-
----
-
-## The Problem
-
-Data pipelines are black boxes. When something goes wrong, you're left asking:
-
-- **"Where did row X go?"** — Dropped somewhere, but which step?
-- **"Why is this value wrong?"** — It was fine in the source, what changed it?
-- **"How did these rows get merged?"** — Which parent records combined?
-- **"Why are there nulls here?"** — When did they appear?
-
-```python
-df = pd.read_csv("customers.csv")
-df = df.dropna()                          # Some rows disappear
-df = df.merge(regions, on="zip")          # New rows appear, some vanish
-df["income"] = df["income"].fillna(0)     # Values change silently
-df = df[df["age"] >= 18]                  # More rows gone
-# What actually happened to customer C-789?
-```
-
-Traditional debugging means `print()` statements, manual diffs, and guesswork. **TracePipe gives you the complete audit trail.**
-
----
-
-## The Solution
-
-```python
-import tracepipe as tp
-import pandas as pd
-
-tp.enable(mode="debug", watch=["income", "score"])
-
-df = pd.read_csv("customers.csv")
-df = df.dropna()
-df["income"] = df["income"].fillna(0)
-df = df.merge(segments, on="customer_id")
-df = df[df["age"] >= 18]
-
-# Pipeline health check
-print(tp.check(df))
-```
-```
-TracePipe Check: [OK] Pipeline healthy
-  Mode: debug
-
-Retention: 847/1000 (84.7%)
-Dropped: 153 rows
-  • DataFrame.dropna: 42
-  • DataFrame.__getitem__[mask]: 111
-
-Value changes: 23 cells modified
-  • DataFrame.fillna: 23 (income)
-```
-
-```python
-# Why did this customer's income change?
-print(tp.why(df, col="income", where={"customer_id": "C-789"}))
-```
-```
-Cell History: row 42, column 'income'
-  Current value: 0.0
-  [i] Was null at step 1 (later recovered)
-      by: DataFrame.fillna
-
-  History (1 change):
-    None -> 0.0
-      by: DataFrame.fillna
-```
-
-**One import. Complete audit trail.**
-
----
-
-## Installation
-
-```bash
-pip install tracepipe
-```
-
----
-
-## Quick Start
-
-### 1. Enable tracking
-
-```python
-import tracepipe as tp
-
-tp.enable(mode="debug", watch=["price", "quantity"])  # Track specific columns
-```
-
-### 2. Run your pipeline normally
-
-```python
-df = pd.DataFrame({
-    "product": ["A", "B", "C", "D"],
-    "price": [10.0, None, 30.0, 40.0],
-    "quantity": [5, 10, 0, 8]
-})
-
-df = df.dropna()                    # Drops row B
-df = df[df["quantity"] > 0]         # Drops row C
-df["total"] = df["price"] * df["quantity"]
-```
-
-### 3. Inspect the lineage
-
-```python
-# Health check - see drops AND changes
-print(tp.check(df))
-```
-```
-TracePipe Check: [OK] Pipeline healthy
-  Mode: debug
-
-Retention: 2/4 (50.0%)
-Dropped: 2 rows
-  • DataFrame.dropna: 1
-  • DataFrame.__getitem__[mask]: 1
-
-Value changes: 2 cells
-  • DataFrame.__setitem__[total]: 2
-```
-
-```python
-# Trace a specific row's full journey
-print(tp.trace(df, where={"product": "A"}))
-```
-```
-Row 0 Journey:
-  Status: [OK] Alive
-
-  Events: 1
-    [MODIFIED] DataFrame.__setitem__[total]: total
-```
-
-```python
-# Explain why a specific cell has its current value
-print(tp.why(df, col="total", row=0))
-```
-```
-Cell History: row 0, column 'total'
-  Current value: 50.0
-
-  History (1 change):
-    None -> 50.0
-      by: DataFrame.__setitem__[total]
-```
-
----
-
-## Key Features
-
-### 🔍 Zero-Code Instrumentation
-
-TracePipe monkey-patches pandas at runtime. Your existing code works unchanged:
-
-```python
-tp.enable()
-# Your existing pipeline runs exactly as before
-# TracePipe silently records everything
-tp.disable()
-```
-
-### 📊 Rich Provenance Data
-
-Track everything that happens in your pipeline:
-
-| Question | Answer |
-|----------|--------|
-| Which rows were dropped? | `tp.check(df)` shows retention by operation |
-| Why did this value change? | `tp.why(df, col="amount", row=5)` shows before/after |
-| What's this row's history? | `tp.trace(df, row=0)` shows full journey |
-| Where did these rows merge from? | Merge parent tracking in debug mode |
-| Which rows grouped together? | `tp.debug.inspect().explain_group("A")` |
-| When did nulls appear? | `tp.why()` flags null introduction |
-
-### 🎯 Business-Key Lookups
-
-Find rows by their values, not internal IDs:
-
-```python
-# Find by business key
-tp.trace(df, where={"customer_id": "C-12345"})
-tp.trace(df, where={"email": "alice@example.com"})
-
-# Find rows where a column is null
-tp.why(df, col="email", where={"email": None})
-```
-
-### 📈 Production-Ready Performance
-
-| Operation | Overhead | Notes |
-|-----------|----------|-------|
-| Filter (dropna, query) | 1.4-1.9x | Acceptable |
-| Transform (fillna, replace) | 1.0-1.2x | Minimal |
-| GroupBy | 1.0-1.2x | Minimal |
-| Sort | 1.4x | Optimized |
-| Scalar access (at/iat) | <1ms added | Fixed overhead |
-
-Tested on DataFrames up to 1M rows with linear scaling.
-
-### 🔒 Safety First
-
-TracePipe never modifies your data or affects computation results:
-
-```python
-# Original pandas method ALWAYS runs first
-# Lineage capture happens after, and failures are non-fatal
-result = df.dropna()  # Guaranteed to work, even if tracking fails
-```
-
----
-
-## Two Modes
-
-### CI Mode (Default)
-Lightweight tracking for production pipelines:
-- Step counts and retention rates
-- Dropped row detection
-- Merge mismatch warnings
-- **No per-row provenance** (fast)
-
-```python
-tp.enable(mode="ci")
-```
-
-### Debug Mode
-Full lineage for development and debugging:
-- Complete row-level history
-- Cell change tracking with before/after values
-- GroupBy membership
-- Merge parent tracking
-
-```python
-tp.enable(mode="debug", watch=["price", "amount"])
-```
-
----
-
-## API Reference
-
-### Core Functions (5)
-
-| Function | Purpose |
-|----------|---------|
-| `tp.enable(mode, watch)` | Start tracking |
-| `tp.check(df)` | Health check with retention stats |
-| `tp.trace(df, row, where)` | Trace a row's journey |
-| `tp.why(df, col, row, where)` | Explain why a cell changed |
-| `tp.report(df, path)` | Export HTML report |
-
-### Control Functions
-
-| Function | Purpose |
-|----------|---------|
-| `tp.disable()` | Stop tracking |
-| `tp.reset()` | Clear all lineage data |
-| `tp.stage(name)` | Label pipeline stages |
-
-### Debug Namespace
-
-For power users who need raw access:
-
-```python
-dbg = tp.debug.inspect()
-dbg.steps              # All recorded operations
-dbg.dropped_rows()     # Set of dropped row IDs
-dbg.explain_row(42)    # Raw lineage for row 42
-dbg.stats()            # Memory and tracking stats
-dbg.export("json", "lineage.json")
-```
-
----
-
-## Data Quality Contracts
-
-Validate your pipeline with fluent assertions:
-
-```python
-result = (tp.contract()
-    .expect_unique("customer_id")
-    .expect_no_nulls("email")
-    .expect_retention(min_rate=0.9)
-    .check(df))
-
-result.raise_if_failed()  # Raises if any contract violated
-```
-
----
-
-## Snapshots & Diff
-
-Compare DataFrame states:
-
-```python
-before = tp.snapshot(df)
-
-# ... transformations ...
-
-after = tp.snapshot(df)
-diff = tp.diff(before, after)
-
-print(f"Rows added: {diff.rows_added}")
-print(f"Rows removed: {diff.rows_removed}")
-print(f"Cells changed: {diff.cells_changed}")
-```
-
----
-
-## HTML Reports
-
-Generate interactive lineage reports:
-
-```python
-tp.report(df, "pipeline_audit.html")
-```
-
-Opens a visual dashboard showing:
-- Pipeline flow diagram
-- Retention funnel
-- Dropped rows by operation
-- Cell change history
-
----
-
-## What's Tracked
-
-| Operation | Tracking | Completeness |
-|-----------|----------|--------------|
-| `dropna`, `drop_duplicates` | Dropped row IDs | FULL |
-| `query`, `df[mask]` | Dropped row IDs | FULL |
-| `head`, `tail`, `sample` | Dropped row IDs | FULL |
-| `fillna`, `replace` | Cell diffs (watched cols) | FULL |
-| `loc[]=`, `iloc[]=`, `at[]=` | Cell diffs | FULL |
-| `merge`, `join` | Parent tracking | FULL |
-| `groupby().agg()` | Group membership | FULL |
-| `sort_values` | Reorder tracking | FULL |
-| `apply`, `pipe` | Output tracked | PARTIAL |
-
----
-
-## Limitations
-
-TracePipe tracks pandas operations, not arbitrary Python code:
-
-| Limitation | Workaround |
-|------------|------------|
-| Direct NumPy array modification | Use pandas methods |
-| Mutable objects in cells (lists, dicts) | Use immutable types |
-| Custom C extensions | Wrap with pandas operations |
-
----
-
-## Example: ML Pipeline Audit
-
-```python
-import tracepipe as tp
-import pandas as pd
-import numpy as np
-
-tp.enable(mode="debug", watch=["age", "income", "label"])
-
-# Load and clean
-df = pd.read_csv("training_data.csv")
-df = df.dropna(subset=["label"])
-df["income"] = df["income"].fillna(df["income"].median())
-df = df[df["age"] >= 18]
-
-# Feature engineering
-df["age_bucket"] = pd.cut(df["age"], bins=[18, 30, 50, 100])
-df["log_income"] = np.log1p(df["income"])
-
-# Audit the pipeline
-print(tp.check(df))
-```
-```
-TracePipe Check: [OK] Pipeline healthy
-  Mode: debug
-
-Retention: 8234/10000 (82.3%)
-Dropped: 1766 rows
-  • DataFrame.dropna: 423
-  • DataFrame.__getitem__[mask]: 1343
-
-Value changes: 892 cells
-  • DataFrame.fillna: 892 (income)
-```
-
-```python
-# Why does this customer have log_income = 0?
-print(tp.why(df, col="income", where={"customer_id": "C-789"}))
-```
-```
-Cell History: row 156, column 'income'
-  Current value: 45000.0
-  [i] Was null at step 1 (later recovered)
-      by: DataFrame.fillna
-
-  History (1 change):
-    None -> 45000.0
-      by: DataFrame.fillna
-```
-
-```python
-# Full journey of a specific row
-print(tp.trace(df, where={"customer_id": "C-789"}))
-```
-```
-Row 156 Journey:
-  Status: [OK] Alive
-
-  Events: 3
-    [MODIFIED] DataFrame.fillna: income
-    [MODIFIED] pd.cut: age_bucket
-    [MODIFIED] DataFrame.__setitem__[log_income]: log_income
-```
-
----
-
-## Benchmarks
-
-Run on MacBook Pro M1, pandas 2.0, Python 3.11:
-
-### Overhead (10K rows, median of 10 runs)
-
-| Operation | Baseline | With TracePipe | Overhead |
-|-----------|----------|----------------|----------|
-| dropna | 0.9ms | 1.7ms | 1.9x |
-| query | 2.1ms | 3.0ms | 1.4x |
-| fillna | 0.4ms | 0.4ms | 1.0x |
-| groupby.sum | 1.2ms | 1.2ms | 1.0x |
-| merge | 4.5ms | 12.6ms | 2.8x |
-| sort_values | 1.1ms | 1.5ms | 1.4x |
-
-### Scale (filter + dropna pipeline)
-
-| Rows | Time | Throughput |
-|------|------|------------|
-| 10K | 5ms | 2M rows/sec |
-| 100K | 35ms | 2.8M rows/sec |
-| 1M | 320ms | 3.1M rows/sec |
-
-### Memory
-
-- Base overhead: ~40 bytes per tracked diff
-- Typical pipeline: 2-3x memory vs baseline
-- Spillover to disk available for large pipelines
-
----
-
-## Documentation
-
-📚 **[Full Documentation](https://gauthierpiarrette.github.io/tracepipe/)**
-
-- [Getting Started](https://gauthierpiarrette.github.io/tracepipe/getting-started/quickstart/)
-- [User Guide](https://gauthierpiarrette.github.io/tracepipe/guide/concepts/)
-- [API Reference](https://gauthierpiarrette.github.io/tracepipe/api/)
-- [Examples](https://gauthierpiarrette.github.io/tracepipe/examples/ml-pipeline/)
-
----
-
-## Contributing
-
-```bash
-git clone https://github.com/gauthierpiarrette/tracepipe.git
-cd tracepipe
-pip install -e ".[dev]"
-
-# Run tests
-pytest tests/ -v
-
-# Run linting
-ruff check tracepipe/ tests/
-
-# Run benchmarks
-python benchmarks/run_all.py
-```
-
-See [CONTRIBUTING](https://gauthierpiarrette.github.io/tracepipe/contributing/) for detailed guidelines.
-
----
-
-## License
-
-MIT License. See [LICENSE](LICENSE) for details.
-
----
-
-<p align="center">
-  <b>Stop guessing where your rows went.</b><br>
-  <code>pip install tracepipe</code>
-</p>